BERT-tiny RAID Detector

This model is used to detect AI-generated texts from the RAID dataset. The model outputs a score corresponding to the likelihood that the text is human.

This model was trained on a stratified subset of RAID's training data.

Model Details

• Base Model: prajjwal1/bert-tiny
• Optimizer: Adam
• Loss Function: FocalLoss
• Hyperparameters:
  • Learning rate: 5e-5
  • Epochs: 5
  • Learning rate: 1e-5
  • Epochs: 5